image credit: wikipedia and PixLoger (from Pixabay)
There is much misunderstanding—perhaps even misdirection among engineers on the topic of dynamic microphone amplification, especially with regard to inline phantom-powered amplifier solutions. This post will attempt to shed some light on this issue.
Microphones can be divided into two categories of the most common types: dynamic and condenser microphones. The two categories of microphones have nearly opposite characteristics in terms of impedance and output level; a condenser capsule has extremely high impedance (that is, its ability to supply current is very limited), such that an amplifier at the capsule is required—but the capsule yields a high output voltage.
Dynamic microphones, in contrast, have low impedance and low output, such that an amplifier is not required in the microphone. In practice, the vast majority of dynamic microphones do not contain amplifiers (save for a few ribbon microphones).
Dynamic microphones can be further classified as moving-coil or ribbon designs. Either design is similar from an electrical perspective; that is, low impedance and low output. In fact, both are so low as to strongly suggest the utility of a step-up transformer inside the microphone, which most dynamic microphones employ.
Dynamic microphone electrical characteristics
First, a brief review of the concepts of signal level and impedance. Signal level is measured by the open-circuit (or unloaded) output of a device, and is either expressed as the microphone’s sensitivity in decibels (most commonly dBV, meaning relative to 1VRMS) or directly as a voltage. Either measure will be for a 1Pa (94dBSPL) signal at the capsule. So, we might find a specification for a dynamic microphone that reads:
Sensitivity: 2mV/Pa (-54dBV/Pa)
We should also find a specification for the microphone’s output or source impedance. Output impedance can be simply thought of as ability to supply current, or instead its capability of driving a load. This figure for a standard professional microphone is nominally 150Ω, and in practice will range between 50Ω and 300Ω. A standard professional microphone amplifier has a nominal impedance of 1.5kΩ, but models may be found ranging from 600Ω to 4kΩ, or even higher. Some models may have a variable input impedance feature that may offer loads down to 150Ω.
The two types of dynamic microphone elements actually have very low source impedance; in the range of a few ohms for moving-coil microphones, and only a small fraction of an ohm for a ribbon element. However, the output from these capsules is lower still than the typical -60 to -50dBV/Pa sensitivity of a complete dynamic microphone.
In fact, their outputs are so low that it would be difficult for any amplifier for directly connect and provide a reasonable signal to noise ratio. Thus, a step-up transformer is placed inside most dynamic microphones to raise their output level, at the cost of an increase in impedance: a transformer is a handy device that lets us swap voltage for current, and vice versa, according to the turns ratio of the two coils of the device. That also means that the impedance ratio is the square of the turns ratio, such that for a 1:4 moving-coil transformer, the output impedance increases by a factor of 16, and for a 1:40 ribbon transformer, by a factor of 1600! But we are content with source impedance in the 150Ω range, and we need the increase in output, so that’s a fair trade.
Dynamic microphone amplification
Standard engineering practice for small-signal audio frequency transmission is to bridge impedance. That means that the load impedance should be greater than the source impedance, in order to minimize signal (voltage) loss. It is opposed to the concept of matching impedance, which maximizes power transfer, which we aren’t concerned with in small-signal amplification (matching impedance is important in radio frequency signal transmission, which is beyond the scope of this article). Generally, it is recommended that the impedance ratio be 1:10 or greater, which is thus realized for the standard microphone and amplifier impedances of 150Ω and 1.5kΩ.
Where the impedances are simple (that is, they do not vary by frequency), we can describe the signal loss for any frequency by this equation:
Vout = Vin * Zload / (Zload + Zsource)
Solving for the standard figures, we have a signal loss of:
Vout = Vin * 1500 / (150 + 1500) = 0.909 Vin
Converting to decibels:
Vout = 20 * log (0.909) = -0.8dB
Quite minimal indeed!
All right then, we have successfully designed a dynamic microphone with reasonable output level and impedance by using a step-up transformer, and connected it to an amplifier of suitable input impedance. Let’s take a moment and look at exactly why it was important to increase its output level with a step-up transformer: noise!
Noise in microphones and amplifiers
Our primary concern in small-signal amplification is noise. All electronic devices have noise, so our job as engineers is to select circuit components that maximize dynamic range by minimizing the accumulation of noise in our audio signal chain. It’s important to understand that all circuits have noise. Anyone who calls a circuit “noise-free” is selling something! So let’s look at our primary noise sources in a dynamic microphone signal chain.
The first source of noise is the microphone capsule itself. In a condenser microphone, that capsule must have an amplifier circuit, so we are given a specification for self-noise of the microphone, which is largely the amplifier circuit’s active noise, but would also include capsule noise. It is given relative to the output of the microphone, as an example:
That indicates that the microphone’s noise is supposedly equivalent to that level of acoustic noise at the microphone’s capsule. It’s a bit complicated whether or not that measure is exactly accurate, especially since there are different weighting factors that can be applied. In that example, A-weighting is used, which is the most generous in terms of yielding a low number for that specification. There are others, such as CCIR quasi-peak, that are intended to be a more accurate comparable measure to acoustic noise, but that’s not important for this article. What is important is that whichever noise measure we use, we can combine that figure with the microphone’s sensitivity to yield an absolute noise figure. Let’s say this is for a condenser microphone that is ten times as sensitive as our given dynamic microphone, that is:
Sensitivity: 20mV/Pa (-34dBV/Pa)
Using this formula:
Noise level = sensitivity – 94dB + self-noise
Noise = -34dBV – 94dB + 18dBA = -110dBA
We get an absolute measure of -110dBA for that’s microphone’s noise floor. As an aside, we’ve mixed units there a little, but that’s okay, because the microphone’s sensitivity is measured at 1kHz, which has a 0dB weighting on the A-weighted scale.
We can take that measure and compare it to our microphone amplifier’s equivalent input noise (EIN) to determine the total noise of our system. The rule-of-thumb for noise is that a following circuit should have at least 6dB, and ideally 10dB, lower noise than a preceding circuit in order to minimize noise accumulation. The formula that determines our accumulated noise, or noise figure, is a bit complicated, but it looks like this:
NF = sqrt (Vnoisemic ^ 2 + Vnoiseamp ^ 2) / Vnoisemic
Did you get 0.5dB for a -120dBA EIN amplifier? Great! And for a -114dBA EIN amplifier: 1.5dB? Outstanding! If not, you’ll have to take my word for it 😉 And if you used a -110dBA EIN amplifier? 3dB, of course!
So we can see that the EIN specification of the amplifier is critical for maintaining a good signal to noise ratio in our signal chain. Fortunately, for condenser microphones, this is generally easy, as most professional microphone amplifiers will have EIN well under -120dBA. The quietest available microphone amplifiers have EIN a few decibels under -130dBA.
At this point, we do need to discuss weighting factors and signal references, because you’ll see variation here. Microphone amplifiers, for reasons of historical interest, are often specified in dBu rather than dBV (which nearly all microphones use). The difference is dBu is 2.2dB higher, don’t forget that! Also, a microphone amplifier might be specified as A-weighted, unweighted, or both, or even something else! Be sure to compare like references! And if EIN is not specified at all, be suspicious!
So if we have a very quiet microphone amplifier of -134dBA, and want to limit noise figure to 1.5dB, we have to make sure that our microphone’s output noise is higher than -128dBA.
But wait, dynamic microphones usually don’t have noise specifications? Are they noise-free? No, they aren’t, they have thermal noise. It can be slightly tricky to figure a dynamic microphone’s noise specification, but a safe estimate is to use its output impedance. Let’s say worst case is 300Ω: the thermal noise equation says (for 10kHz bandwidth, which approximates A-weighting, and standard room temperature):
Ntherm = sqrt (4 * 1.38 * 10^-23 * 300K * 300 Ω) * sqrt (10kHz) = -133dBA
I am tired from all of that math, let’s take a break!
Okay, we are back. So we can see that even if we found a magical noise-free microphone amplifier, we are still stuck with about the same level of thermal noise. We can nip at that a bit by using the best possible lowest-resistance microphone transformers we can find, and we’ll gain a few decibels, but that’s about all we can do.
So, if we use -134dBA as our best-possible absolute noise floor, we just have to compare that to our microphone’s sensitivity to figure effective self-noise:
Self-noise = 94dB – (-54dBV/Pa) + (-134dBA) = 14dBA
Hey, not too bad, if we do everything *perfectly*, we get self-noise comparable to a condenser microphone.
Noise and gain
It is often said that dynamic microphones, when used on quiet sources (<94dBSPL), require “a lot of clean gain”. Essentially true, but how do we quantify that? There are two statements made “clean” and “gain”, which encompass three characteristics of amplifiers: noise, distortion, and gain. It’s entirely possible to design an amplifier that has a lot of gain, especially if we choose a circuit topology such as an operational amplifier (“opamp”). A typical specification for an opamp will yield something on the order of 120dB+ gain! We don’t need that much, hopefully! And in fact, it’s common to see a general purpose microphone preamplifier with maximum gain on the order of +40-60dB. Higher gain figures may be available from specialized dynamic microphone preamplifiers. However, we should ask, does anyone ever really need more than +60dB of gain from a preamplifier? For a recording application, probably not: we have already figured that our absolute noise floor is around -134dBA; that plus +60dB is -74dBA. What device are we feeding that has a higher noise floor than that? If we are using an analog-to-digital converter (ADC) that has 0dBFS = +19dBu (15dB of headroom above +4dBu, a typical specification), which is +17dBV, that means that a noise floor of -84dBA (10dB below our incoming noise floor) yields dynamic range of only 101dB. That’s not a very good specification for a modern ADC, even in a portable unit! More typically, we would expect at least 110dB dynamic range. Clearly, 60dB gain is more than enough gain at a preamplifier for any microphone to feed an ADC.
That said, it can certainly be convenient to have more gain, especially if trying to match levels with hotter-output microphones, or if directly supplying a power amplifier. But let’s be careful not to seek more gain that is necessary, merely in pursuit of a higher peak signal level that doesn’t actually have any advantage in dynamic range. This is because increasing gain in a single stage can have negative effects, depending on the circuit design: possibly a reduction in bandwidth and an increase in distortion. Designers seek to avoid that, so they will limit gain in order to preserve performance, where indicated, often by dividing gain into two or more amplifier stages.
There is another misconception about preamplifier noise and gain; that is, that preamplifiers are noisier at higher gain settings. In fact, the opposite is usually true: EIN is usually lowest at maximum gain. It may be easier to hear the noise floor, but that’s only because the signal is much louder too. This occurs because many preamplifier designs have more than one stage, with the latter stage(s) being noisier than the input gain stage. The presumption in the design is that the higher EIN figure at lower gain doesn’t matter, because the incoming signal will be hot enough to swamp the preamp’s noise. So we don’t usually have to worry about using the higher gain settings on a preamplifier, if it is at all a reasonable design.
In the real world
It’s not always that easy to find a very-low noise microphone amplifier, for example with low-to-midrange mixers, and especially with portable devices. We might often find that such devices have EIN that is 10dB or more worse than ideal. Some might be 20dB noisier! How can we cope if we need to record very quiet sources in low (acoustic) noise environments?
I would suggest that the easiest course is to use a condenser microphone with low self-noise and high output, which eliminates the requirement for a very low-noise amplifier. But then I’m biased, and you didn’t come to this article because you want to use a condenser microphone. You have selected a dynamic microphone for your application, and you need it to be quiet. What to do?
Inline amplifier solutions—and problems!
It may not be feasible for reasons of cost, size, or power consumption to use a very low-noise dedicated dynamic microphone amplifier. Being aware of these factors, the market has provided a range of inline amplification solutions to address this need. Naiant has supplied such solutions for many years with its PFA, IPA, and now the new IFA range (note: the IFA has replaced the PFA for dynamic microphone amplification applications).
However, I have grown concerned that certain other solutions on the market are being promoted as superior when in fact they have design deficiencies that could negate any claimed product benefits.
Let’s first describe what an inline dynamic microphone amplifier should do; it should have very low equivalent input noise (I would suggest -127dBA as a maximum, but ideally lower), relatively high load impedance (>1.5kΩ, preferably higher), and just as critically, low output impedance (no more than 150Ω). It should provide +20dB or more gain, in order to bring its output noise well above the level of even the most questionable following microphone preamplifiers. Its output should be well balanced and it should have either a differential input stage, or its gain on each input leg should be carefully matched such that the following stage will cancel the entire signal chain’s interference.
We have already discussed input noise and load impedance; gain is straightforward to understand from that point. Output impedance of the inline amplifier is not specified by all manufacturers, but it is just as important, especially since we assume this device will enable us to drive long cables without loss. It should at least be as good as the microphone itself at that task—but we want better, right? Doesn’t a high signal level allow us to drive longer cables? Maybe; maybe not.
Care and feeding of microphone cables
As stated earlier, output or source impedance can be thought of as the ability to drive a load. Thus far, we have only considered the load of the amplifier, and only as a simple resistive (that is, not frequency dependent) load. We must also consider the load of the microphone cable, which, for our purposes, is primarily capacitive (the resistance of any proper microphone cable is negligible compared to the other impedances in the system, so we can ignore it).
The source impedance of a device connected to a cable forms a simple, single-pole RC filter, which is a lowpass filter. According to this equation, frequencies above the corner will be attenuated by 6dB/octave:
Corner freq = 1 / (2 * pi * Ω * C)
Where C is capacitance in farads. A good quality microphone cable will have capacitance of about 100pF/m, and a standard professional microphone will thus be capable of driving a 100m cable without trouble:
1 / ( 2 * pi * 150 Ω * 100pF/m * 100m) = 106kHz
Great, that’s far above our 20kHz audio bandwidth. So as long as our inline amplifier also has 150Ω source impedance, no worries there either!
But what if our inline device has inappropriately high output impedance, say 3kΩ? Our bandwidth with our 100m cable is only 5.3kHz!
In using long cables, we are also rightly concerned with interference. Balanced signal transmission is really good at cancelling interference, but in order for that to work, two things must be true: the two legs of the output device must have well matched impedance (a matched, opposite polarity signal is not important, in fact it is irrelevant—there can be no signal at all on one leg, so long as the impedance matches) *and* the receiving device must either be a differential input device with good common-mode rejection ratio (CMRR), or, if we know that the device following that is such a device, our interim inline device may merely have well-matched input and output impedance, and also well-matched gain.
If we fail at any of those requirements, our inline amplifier device may have worse induced interference than if we hadn’t used it, maybe making overall signal-to-noise worse than without any inline amplifier. Are there such devices on the market? I fear there are.
Now, if you have such a device, you may either use it with short enough cables to avoid any such problems (I would recommend less than 10m), or you may try using it at the amplifier end of the cable, rather than the microphone end—that will fix the bandwidth problem, and may help with any interference problems. Or it may not. Caveat emptor!
Inline amplifier design
Wait, you ask, why are there any such devices on the market? Isn’t it possible to design a proper inline amplifier? I can only guess at the first question, but the answer to the second is a clear affirmative.
So let’s speculate on the first question. I have noted that some manufacturers promote the junction field-effect transistor (JFET, or often simply FET) as the proper solution to this circuit requirement. I openly question whether that is the case. The other common type of transistor used for small-signal audio-frequency amplification is the bipolar junction transistor, or BJT (there are also MOSFETs sometimes used in audio circuits, but we’ll ignore them for now as they are uncommon in this application). The two types have distinguishing characteristics; it is not a case of one or the other simply being better, but rather, which is best for a given function in a circuit. Many circuits will use both!
Simply, a JFET can be thought of as a voltage-input device; it has an extremely high input impedance, and so is used in applications where source impedance is high: instrument amplifiers and the capsule amplifier in condenser microphones to name a couple. Compared with the same grade of BJT, a JFET will have lower gain and higher noise. Yes, higher noise! Don’t take my word for it; if you don’t want to read a bunch of datasheets, here are a couple of pages that reference BJT and JFET characteristics:
“The 2SK170 and some other high gm fets are about 10 times the price of good low noise bjts. Compared to the 2SC2547 bjt the 2SK170 fet has typically twice the noise voltage, twice the feedback capacitance, and a 5th of the gm at 3mA.”
https://www.embedded.com/print/4015906 (defunct link)
“For a typical JFET operating at 2 mA drain current, the gfs value will be of the order of 1″4 mS, which would give a stage gain of up to 40 if R2, in Figure 9.13, is 10 kΩ. This is very much lower than would be given by a BJT and is the main reason why they are not often used as voltage-amplifying devices in audio systems unless their very high input impedance (typical values are of the order of 1012 Ω) or their high, and largely constant, drain impedance characteristics are advantageous.”
So, why use a JFET for an inline dynamic microphone amplifier if its noise is higher and gain is lower than a comparable BJT? It’s not because of input impedance; it’s not hard to get a BJT circuit with input impedance much greater than the 3kΩ we want for our inline amplifier. 30kΩ, even 100kΩ; not a problem. We certainly don’t need the 10MΩ+ a JFET can manage. It seems to be a case of trading away something we really need (low noise, high gain) for something we don’t (high input impedance).
So again, why does anyone do this? My best guess is because it’s easy; a simple phantom-powered JFET circuit can be made point-to-point in a small space using only four components: two JFETs and two resistors. This will be a fairly terrible amplifier: it’s not differential input, gain doesn’t match between the two legs (because JFETs are notoriously not well matched for gain), headroom will be low due to use of “self-bias” at the input and no feedback, gain is a function of the input impedance of the following device, and output impedance is equal to the input impedance of the following device.
This JFET amplifier topology is called common source. Note the section on characteristics from this wikipedia entry:
Common Source Amplifier Characteristics
The output impedance characteristic can be huge problem if you are trying to drive a long cable, and it gets worse as you increase the input impedance of the following device (your microphone preamplifier). One manufacturer of an inline device shows a graph of gain vs. following input impedance (a strong hint at the amplifier topology), and recommends using higher preamplifier input impedance to maximize gain. But we have seen above what the effect on bandwidth and possibly induced interference will be from higher impedance on a long cable run. This is simply not proper engineering practice! Another manufacturer promotes the purity of their signal path as only consisting of JFETs—as if BJTs, resistors, and capacitors are great evils than distortion, noise, interference, and lack of adequate bandwidth. And in all likelihood, all of those supposedly-verboten components will rightly be used in the microphone preamplifier anyway.
It’s possible to make that simple JFET circuit better by adding a few more resistors (maybe to improve bias, or to make a differential input), by adding another JFET stage for more gain, by paralleling JFETs to reduce noise, by matching JFETs for gain, by using feedback, etc. But at some point the designer must ask why all the trouble versus using a standard high-gain, low-noise, differential input BJT circuit that has been well described and published 40+ years ago? And I honestly can’t answer that. I don’t think it’s a matter of cost, because either part—JFET or BJT—costs less than $0.20 in quantity. It must then be a matter of convenience in assembly, or just marketing-speak to move product.
I do recommend that you ask every inline device manufacturer for these specifications before you buy: input impedance, output impedance, gain, maximum input level (at a given level of THD), EIN, CMRR (if relevant). If they won’t or can’t provide those, think very hard about your signal integrity as it passes through that device, because they may not have thought too much about that in their design.